IT pros at Microsoft's annual IT confab next week hope to see a beta for the first service pack for Microsoft Windows Server 2008 R2 and Hyper-V R2, the latter of which is expected to add support for memory overcommit and the Remote FX desktop virtualization protocol.
Microsoft would not confirm timing for the first service packs (SP1) for Windows Server 2008 R2, Windows 7 or Hyper-V R2, but TechEd 2010 is the company's usual launching pad for feature updates and new products, so customers said they expect to see at least an SP1 beta for those products.
Microsoft previously said the Exchange Server 2010 SP1 will launch at TechEd with improvements to archiving and discovery, Outlook Web App (OWA), mobile user and management improvements, and some additional UI for management tasks.
Many people consider the first service pack to be the maturity point when it is safe to upgrade to a new product, so Microsoft wants to deliver a beta at TechEd, said Michael Cherry, an analyst with Kirkland, Wash.-based Directions on Microsoft.
What to expect from Hyper-V R2 SP1
Perhaps the biggest improvements coming are for Hyper-V R2, which launched last year in September with live migration and 64-processor support. From a competitive standpoint, however, it was considered incomplete without memory management, which VMware Inc. has offered in its hypervisor for years.
Up until recently, Microsoft maintained it isn't a good idea to over-provision memory resources, but in reality, memory overcommit is a valuable feature comparable to pagefile in the Windows OS, and users demanded it. Microsoft conceded to the value of memory overcommit when it unveiled its own version, Dynamic Memory, in March.
"It's hard to make a case that it isn't important for virtualization when the Windows OS has used memory paging to overcommit memory for Windows services and applications for more than a decade," said Chris Wolf, a Burton Group virtualization analyst.
Hyper-V R2 user Robert McShinsky, a senior systems engineer for Dartmouth Hitchcock Medical Center in Lebanon, N.H., said memory is always the first resource to run out, and it is expensive to add.
For instance, he typically orders host servers with 8 GB DIMMs, 12 DIMM slots and 96 GB of memory. That box costs about $12,500, with the RAM accounting for $8,000 to $9,000 of that cost, and the price goes up exponentially for 16 GB DIMMs.
He estimates with the cost of host hardware, host OS and support software, and SAN resources and connectivity, the average virtual slot costs around $1,000. So, if the Dynamic Memory feature increases usable RAM and allows four extra virtual machines (VMs) per host, the cost savings could be $4,000 per host, McShinsky said.
Microsoft claims it designed Dynamic Memory for production environments, but some users say it will be particularly beneficial for testing and development.
"[Test and development] environments are the areas that suffer from VM bloat the most and could benefit most from a little extra headroom," McShinsky said. "This would translate into direct cost savings, a pat on the back and I am sure a huge raise. Well, maybe not those last two."
Here's what's in Hyper-V SP1
The Midvale, Utah-based independent analyst firm Burton Group reported that Hyper-V R2 was not ready for production when it came out last year, and while Dynamic Memory is a good start, there's room for improvement, Wolf said.
For example, in-guest memory counters include Dynamic Memory reservations. This means if a guest is using 1 GB of memory but also has a 500 MB reservation, system monitor counters and task manager make it look like the VM is consuming 1.5 GB of RAM, Wolf said.
"It may be a small point, but capacity management and performance monitoring tools may initially report inaccurate guest-level memory consumption and requirements," Wolf said. "This will occur until these tools properly account for Hyper-V Dynamic Memory reservations when collecting and reporting performance data."
In SP1, Hyper-V R2 will support Linux SMP, but the hypervisor will not offer logical VM prioritization. Ideally, administrators should be able to set VM priorities so the most critical VMs are restarted first following a physical server failure, Wolf said.
A virtual machine priority restart feature is also critical when a multi-tier application stack consumes several VMs and requires a specific VM start order (for example, the back-end database is online before dependent front-end applications). Hyper-V R2 allows automatic or manual restart settings, but the latter means an administrator has to bring VMs back online following a failure.
"We do not consider that adequate for production applications," Wolf said.
Growing up with Hyper-V
Dave Sobel, a Microsoft partner and CEO of Evolve Technologies, said that despite some shortcomings, Microsoft shops that have adopted Hyper-V R2 can rest assured that improvements will keep rolling in.
"I have no problem rolling out Hyper-V R2 to my customers as it is now. It is pretty solid, and knowing that incremental feature improvements come in the form of service packs helps," Sobel said. "The ROI on Hyper-V is very compelling for Windows Server users, so to see them going head to head with VMware features is fantastic for customers."
Wolf said that for an organization heavily vested in Microsoft's management stack, growing with Hyper-V R2 today rather than migrating to it later is a sensible choice.